Depression　is　a　common　psychological　disorder,　which　can　be　extremely　damaging　people　both　physically　and　psychologically.　A　rapid　and　accurate　programme　of　review　of　depressed　mood　is　therefore　important.　This　project　used　the　Depression　Rest　dataset　for　2　classifications　of　depression,　including　none　and　depression.　A　hybrid　CNN-LSTM　network　was　used,　combining　the　advantages　of　convolutional　neural　networks　and　long　and　short-term　memory　networks,　which　is　good　for　processing　both　sequential　and　temporal　data.　After　tuning　and　training,　the　test　set　achieved　an　average　classification　accuracy　of　95.5%　with　excellent　performance.　The　research　significance　of　this　project　is　that　depression　is　a　common　psychological　disorder,　but　it　is　difficult　to　diagnose　with　subjective　observation.　Big　data　analysis　and　machine　learning　techniques　to　accurately　determine　the　degree　of　depression　can　provide　strong　support　for　early　screening,　diagnosis　and　treatment　of　depression.　The　study　also　provides　a　valuable　reference　for　future　automated　testing　of　depression　levels　based　on　biomarkers.　©　2023　IEEE.